When to use Agents for Engineering Workflows
Use these agents when you need to automate engineering tasks that are:- Repetitive (provisioning, deployments, audits, checks).
- Standardized (pipeline operations, resource creation, validations).
- High-context (logs, metrics, environments, IaC).
- Triggered by users, workflows, chat commands, or webhooks.
Prerequisites
Before creating these agents ensure:- At least one Environment is Ready
- That Environment has a Worker Queue with at least one connected worker
- Required secrets exist (cloud credentials, registry tokens, access keys)
- Required Skills are enabled at the Environment level (shell, Docker, Python, etc.)
CloudOps Engineer Agent
A CloudOps Engineer agent manages cloud infrastructure operations. It handles tasks such as provisioning, resource audits, monitoring, and deployment execution across providers such as AWS, GCP, or Azure.What this agent does
A CloudOps Engineer typically performs:- Cloud resource inspections and audits
- VM, Kubernetes, or serverless deployment actions
- Reading logs and metrics to surface anomalies
- Managing cloud accounts/tokens stored in Environments
- Responding to alerts or executing runbooks autonomously
When to use it
Choose this agent when you need:- Multi-cloud operational support
- Automated cloud infrastructure checks
- On-demand remediation (restart nodes, scale resources)
- Daily/weekly cost or resource drift audits
- Integration with alerting workflows (Slack, webhooks)
Required configuration
This agent requires:Environment-level prerequisites
- Cloud credentials (AWS IAM key, GCP service account, Azure SP)
- Skills:
- Shell (Full Access or Restricted)
- Docker (if running local tooling)
- Python (Unrestricted or Restricted)
- Kubiya Platform APIs enabled
Optional MCP integrations
- Cloud provider-specific MCP servers
- GitHub/GitLab for IaC repos
- Terraform Cloud/Atlantis MCP
Step-by-step: Create a CloudOps Engineer agent
1. Basic Info
- Name: CloudOps Engineer
- Description: Manages cloud infrastructure. Performs audits, deployments, and operational diagnostics.
- Model: Claude Sonnet or Claude Opus for high-precision operational tasks
- Capabilities: cloudops, infrastructure, devops
2. Deployment
- Environments: Add a production, staging, or shared cloud environment
- Runtime: Default recommended
3. Execution Environment
Attach:- Environment variables such as
DEFAULT_REGION=us-east-1 - Secrets like
AWS_ACCESS_KEY,AWS_SECRET_KEY,GCP_CREDENTIALS - Integrations: GitHub if working with IaC repos
4. Tools (Skills & MCP)
Skills to enable:- Shell – Full Access
- Python – Unrestricted
- Docker – Full Control
- Cloud provider MCP (aws-cli, gcloud, az)
- Terraform MCP
5. Policies (after creation)
Example restrictions:- Restrict shell to investigative commands only
- Allow deployments only in specific namespaces or regions
How to use the CloudOps Engineer
Once created, you can:Chat-based
Ask: “Check unused EC2 instances older than 30 days.” “Run a drift check against staging Terraform.” “Analyze CloudWatch errors for service X.”Workflow-based
Use this agent in automation:- Daily cost audits
- Auto-remediation tasks
- Alert enrichment
- Deployment orchestration
Webhook-based
Trigger from cloud alerts or monitoring tools. You can verify your agent by running the provided command with your agent ID:Platform Engineer Agent
A Platform Engineer agent manages platform infrastructure, CI/CD tooling, internal developer platforms, and pipeline orchestration. It is ideal for automating internal platform tasks.What this agent does
A Platform Engineer agent typically performs:- Pipeline debugging and execution
- Infrastructure-as-code workflows (Terraform, Helm, Argo, Flux)
- Container builds and validations
- Deployments across clusters/environments
- Reading logs and analyzing build failures
- Managing platform lifecycle automation
When to use it
Use this agent when you need:- Developer workflow automation
- Consistent CI/CD operations
- Standardized environment/platform management
- Automated validation of IaC or manifests
- Assistance with Kubernetes, Helm, Docker, Terraform tooling
Required configuration
Environment-level prerequisites
- CI/CD credentials
- Kubernetes API tokens
- Registry credentials
- Skills:
- Shell – Full Access
- Docker – Full Control
- Full Diagramming Suite (optional for workflow visualization)
Optional MCP integrations:
- GitHub/GitLab
- ArgoCD MCP
- Kubernetes MCP
- Jenkins MCP
Step-by-step: Create a Platform Engineer agent
1. Basic Info
- Name: Platform Engineer
- Description: Automates platform tasks, pipelines, and deployments.
- Model: Claude Sonnet for performance or Claude Opus for deeper reasoning
- Capabilities: devops, platform, cicd
2. Deployment
- Environments: staging, prod, infra
- Supports multi-environment use
3. Execution Environment
Add:K8S_CONTEXT=prod-cluster- Registry secrets
- GitHub integration for pipeline repos
4. Tools (Skills & MCP)
Recommended Skills:- Docker – Full Control
- Shell – Full Access
- Full Diagramming Suite
- Python (optional)
- Kubernetes
- ArgoCD
- GitHub Repos
5. Policies
Define guardrails:- Only deploy to specific namespaces
- Allow Docker operations only on controlled repos
- Restrict shell commands
How to use the Platform Engineer
Chat-based workflows
Ask the agent to: “Validate this Helm chart and show me errors.” “Deploy service X to staging.” “Explain why pipeline run #438 failed.”Workflow-based automation
Attach it to:- PR-triggered IaC validations
- Continuous deployment pipelines
- Automatic container scanning and fixes
Developer workflow self-service
Developers can delegate tasks via chat: “Create a temporary Kubernetes namespace for testing.” Run the provided cli command to verify your new agent:Troubleshooting Both Agents
If CloudOps or Platform Engineer agents appear stuck:- Ensure the Environment is in Ready state
- Confirm the Worker Queue has at least one active worker
- Check that required Skills are attached
- Verify secrets and credentials exist at the correct level
- Review OPA policies for command or environment restrictions
Best Practices
- Keep Skills minimal; add more only when needed
- Store cloud tokens and CI credentials at the Environment level
- Use multiple Environments instead of duplicating agents
- Add policies once the agent’s behavior is verified
- Always start with safe read-only commands during testing
Next Steps
- Add alerts or workflows that trigger these agents
- Connect MCP integrations for deeper automation
- Build dashboards showing agent health and activity
- Add environment-specific OPA policies for guardrails